Re: pkg/57074: sysutils/pv fails to build on modern MacOS, tries to use stat64

[dns%strangeloop.cc@localhost]

On 2022-10-27 01:15, David Holland wrote:
> The following reply was made to PR pkg/57074; it has been noted by 
> GNATS.
>
> From: David Holland <dholland-pbugs%netbsd.org@localhost>
> To: gnats-bugs%netbsd.org@localhost
> Cc:
> Subject: Re: pkg/57074: sysutils/pv fails to build on modern MacOS, 
> tries to
>  use stat64
> Date: Wed, 26 Oct 2022 22:13:16 +0000
>
>  On Tue, Oct 25, 2022 at 04:25:00PM +0000, dns%strangeloop.cc@localhost wrote:
>   > .if ${OPSYS} == "Darwin" && ${MACHINE_ARCH} == "aarch64"
>   > CONFIGURE_ENV+=         ac_cv_func_stat64=no
>   > .endif
>   >
>   > Will this break build on other Darwin's or older Mac's? Maybe
>   > checking OPSYS_VERSION > 120000 is a better deal? Unfortunately I
>   > don't know much about Mac's and my system is currently 120600.
>
>  It might. The whole sorry history of this stat64 nonsense has a lot of
>  tentacles. But probably not since aarch64 MacOS doesn't go back very
>  far.
>
>  I guess the other question is: why does the configure test come up
>  with the wrong answer? If we can figure out what's up with that, it's
>  a more robust way to fix things. config.log may provide some insight.
>
>  --
>  David A. Holland
>  dholland%netbsd.org@localhost

I tried my best and I did eventually came up with this diff, which 
should explain things. I included a few more lines to show how the 
macros work:

--- autoconf/header.in.orig     2021-09-04 22:59:47.000000000 +0300
+++ autoconf/header.in  2022-10-28 00:02:23.000000000 +0300
@@ -58,21 +58,21 @@
 #ifdef ENABLE_LARGEFILE
 # define __USE_LARGEFILE64 1
 # define _LARGEFILE64_SOURCE 1
 #else
 /*
  * Some Macs have stat64 despite not having open64 while others don't 
have
  * either, so here even if we don't have open64 or LFS is disabled, we 
have
  * to check whether stat64 exists and only redefine it if it doesn't
  * otherwise some Macs fail to compile.
  */
-# ifdef __APPLE__
+# if defined(__APPLE__) && !defined(_DARWIN_FEATURE_64_BIT_INODE)
 #  ifndef HAVE_STAT64
 #   define stat64 stat
 #   define fstat64 fstat
 #   define lstat64 lstat
 #  endif
 # else
 #  define stat64 stat
 #  define fstat64 fstat
 #  define lstat64 lstat
 # endif


I could not easily get around the detection of HAVE_STAT64 (autoconf 
defines this from successful compilation of function stat64(2)) but by 
using this "64_BIT_NODE" macro, defined on modern (Mac OSX 10.5 or 
higher) systems, this will ignore HAVE_STAT64 and fallback to the 
default behaviour of other OS's, which is to "define stat64 stat" and 
hence fix the code.

So, until Apple removes stat64(2) altogether this should work.. :)

Cheers,
Dennis